WO2014171618A1

WO2014171618A1 - Storage system and method for processing writing data of storage system

Info

Publication number: WO2014171618A1
Application number: PCT/KR2014/000867
Authority: WO
Inventors: 이재수
Original assignee: 주식회사 디에이아이오
Priority date: 2013-04-17
Filing date: 2014-01-29
Publication date: 2014-10-23
Also published as: KR20140124537A; US20160162187A1; KR101478168B1

Abstract

In a method for processing writing data of a storage system, when a file system makes a writing request, writing data related to the writing request is stored in a data input and output queue, all data stored in the data input and output queue is classified into a complete storage logic page and an incomplete storage logic page, data corresponding to the complete storage logic page is transmitted to a storage device in response to the writing request, and data corresponding to the incomplete storage logic page remains in the data input and output queue in response to the writing request.

Description

How storage systems and write data are handled by storage systems

The present invention relates to a semiconductor memory system, and more particularly, to a storage system comprising a host device and a storage device (eg, a flash memory device) and a method of processing write data thereof.

The semiconductor memory device may be classified into a volatile memory device and a nonvolatile memory device according to whether data can be stored in a state where power is not supplied. Recently, according to the trend of miniaturization and large capacity of semiconductor memory devices, NAND flash memory devices (NAND flash memory device) among nonvolatile memory devices are widely used for miniaturization and large capacity, and storage devices in the form of NAND flash memory devices (eg For example, solid state drives (SSDs) are replacing hard disk drives (HDDs). In this case, the storage device may include at least one NAND flash memory and a storage controller for controlling the same, and the host device may include a file system and a host controller for performing an interaction with the storage controller. It may include.

Recently, a read operation, a write operation, an erase operation, a merge operation, a copyback operation, a compaction operation, and a garbage collection of a NAND flash memory are performed. Attempts to distribute the functionality of the flash translation layer (FTL) that supports the filesystem to the storage device and the host device by controlling operations, wear leveling, address mapping operations, etc. It is done. For example, the host controller may include a host flash translation layer, and the storage controller may include a storage flash translation layer.

In this case, the basic units of read and write operations used by the host device are block units (for example, 1 KB, 2 KB, 4 KB, etc.), and the host logical page size used by the host device is determined by considering the block size of the file system. Set it. On the other hand, the basic unit of read and write operations used by the storage device is a multiple of the physical page of flash memory (for example, 16 KB or larger), which has recently improved the storage device's continuous write performance and reduced the mapping table size. For example, the logical page size of a storage device is increasing. Accordingly, the size of the storage logical page tends to be larger than the size of the host logical page.

As such, as the size of the storage logical page becomes larger than the size of the host logical page, writes to the same storage logical page are frequently requested as a plurality of writes. This is a problem in that the overwrite is not supported and the performance of the flash memory is greatly reduced and the life is shortened due to the characteristics of the flash memory that performs read operation, write operation in page units, and erase operations in block units.

An object of the present invention is to provide a write data processing method of a storage system capable of preventing a write operation from being repeatedly performed multiple times on the same storage logical page in a storage device.

Another object of the present invention is to provide a storage system employing the write data processing method.

However, the object of the present invention is not limited to the above objects, and may be variously expanded within a range without departing from the spirit and scope of the present invention.

In order to achieve the object of the present invention, a write data processing method of a storage system according to an embodiment of the present invention, when the file system writes a request for one or more blocks, the host controller and the write request; The write data is stored in a data I / O queue, and all data stored in the data I / O queue are divided into full storage logical pages and incomplete storage logical pages. The data transferred to the storage device and corresponding to the incomplete storage logical page are merged with the next write data to remain in the data input / output queue so as to become a complete storage logical page and wait for the next write data.

According to an embodiment, when the write data includes the complete storage logical page, data corresponding to the complete storage logical page may be transmitted to the storage device without being stored in the data input / output queue.

According to an embodiment, when the write data does not include data of a storage logical page corresponding to residual data associated with a previous write request stored in the data input / output queue, the remaining data may be completely stored even if the residual data is further retained. Since it is unlikely to be a logical page, the residual data may be transmitted to the storage device as an incomplete logical page.

In example embodiments, the complete storage logical page and the incomplete storage logical page may be classified based on a storage logical page size set in the storage device.

According to an embodiment, when the file system requests a flush, all data stored in the data input / output queue may be transmitted to the storage device.

In order to achieve another object of the present invention, a storage system according to embodiments of the present invention may include a storage device and a host device. The storage device may include at least one flash memory and a storage controller to control the flash memory. The host device may include a file system and a host controller that interacts with the storage controller. At this time, when the file system makes a write request, the host controller stores write data associated with the write request in a data input / output queue, and stores all data stored in the data input / output queue in a complete storage logical page and incomplete storage. The data corresponding to the complete storage logical page may be transmitted to the storage device by dividing into logical pages.

According to an embodiment, when the write data includes the complete storage logical page, the host controller may transmit data corresponding to the complete storage logical page to the storage device without storing the data corresponding to the complete storage logical page.

According to an embodiment of the present disclosure, when the write data does not include data of a storage logical page corresponding to residual data associated with a previous write request stored in the data input / output queue, the host controller stores the residual data in the storage. Can be sent to the device.

In example embodiments, the host controller may classify the complete storage logical page and the incomplete storage logical page based on a storage logical page size set in the storage device.

According to an embodiment, when the file system requests a flush, the host controller may transmit all data stored in the data input / output queue to the storage device.

Recently, as the logical page size of flash memory increases, the size of the storage logical page tends to be larger than that of the host logical page. As a result, a write request for the same storage logical page may be requested as a plurality of writes in response to a write request of the file system.

If the size of the storage logical page is larger than the size of the host logical page, using the storage system and the write data processing method of the storage system according to an embodiment of the present invention, the write data associated with the write request is provided with a data input / output queue in the host controller. Can be transferred in units of storage logical page sizes. This can reduce unnecessary transfer of write requests and can significantly improve the performance of the storage system and increase the lifespan of the storage system by reducing the number of full merges during garbage collection processing in the storage FTL.

However, the effects of the present invention are not limited to the above effects, and may be variously extended within a range without departing from the spirit and scope of the present invention.

1 is a block diagram illustrating a storage system according to example embodiments.

2A illustrates an example in which a host controller operates in response to a write request of a file system in a conventional storage system.

FIG. 2B is a diagram illustrating an example in which the host controller operates in response to a write request of a file system in the storage system of FIG. 1.

FIG. 3 is a diagram illustrating an example in which a host controller operates in response to a flush request of a file system in the storage system of FIG. 1.

4 is a flowchart illustrating a write data processing method of a storage system according to example embodiments.

Hereinafter, with reference to the accompanying drawings, it will be described in detail a preferred embodiment of the present invention. The same reference numerals are used for the same elements in the drawings, and duplicate descriptions of the same elements are omitted.

Referring to FIG. 1, the storage system 100 may include a host device 120 and a storage device 140.

The storage device 140 may correspond to a NAND flash memory device, and may include a solid state drive (SSD), a secure digital card (SDCARD), a universal flash storage (UFS), and an embedded device. It may be implemented as an embedded multi media card (EMMC), a compact flash card, a memory stick, an XD picture card, or the like. However, this is merely an example, and the type of the storage device 140 is not limited thereto. The storage device 140 may include first to n th (where n is an integer of 1 or more) flash memories 146-1,..., 146-n and a storage controller 142 that interact with each other. Can be. The storage device 140 supports the file system 122 by the storage controller 142 using the flash translation layer, thereby reading, writing, erasing, merging, copyback, compaction, and garbage collection. , Wear leveling operations, and the like. Although not shown in FIG. 1, the storage device 140 may have hardware or software other than the first through nth flash memories 146-1,..., 146-n and the storage controller 142. It may further include components.

The host device 120 may include a file system 122 and a host controller 124 that interact with each other. In this case, the file system 122 generates I / O commands to the host controller 124 in block units (hereinafter, referred to as a host logical page), and the host controller 124 is in sector units (for example, 512 bytes) may generate I / O commands for the storage device 140 to process the I / O commands of the file system 122.

The host controller 124 may include a data input / output queue 126. The data input / output queue 126 is a temporary storage space for collecting and processing consecutive write requests from the file system 122 when they are related to the same storage logical page. The meaning of the data input / output queue 126 is temporary storage space. It should be construed as including all data structures (e.g., linked lists, priority queues, etc.) that function as a. If the size of the storage logical page is larger than the host logical page, if writes to the same storage logical page are sent in multiple write requests, the overall performance of the storage system 100 may be degraded and the lifetime may be shortened. The queue 126 may serve to compensate for this.

In addition, the host controller 124 may divide all data stored in the data input / output queue 126 into a complete storage logical page and an incomplete storage logical page. In this case, if all data of each storage logical page stored in the data input / output queue 126 or temporary storage space is completely, the storage logical page is called a full storage logical page. We define a case as a partial storage logical page. In one embodiment, whether or not the complete storage logical page is a complete storage logical page if it contains data corresponding to the last portion of the storage logical page, otherwise it may be determined as an incomplete storage logical page. In one embodiment, the host controller 124 compares the storage logical page size with the size of data corresponding to the same storage logical page in the data I / O queue 126, and determines that the storage logical page is a complete storage logical page if the sizes are the same. If the size of the data determined to correspond to the same storage logical page in the data I / O queue is smaller than the storage logical page size, it may be determined to be an incomplete storage logical page.

When a complete storage logical page exists in the data input / output queue 126, the host controller 124 may transmit data corresponding to the complete storage logical page to the storage device 140. On the other hand, data corresponding to an incomplete storage logical page in the data input / output queue 126 may remain in the data input / output queue 126 and wait for the next write request. If data corresponding to an incomplete storage logical page remains in the data input / output queue 126 and waits for the next write request, data corresponding to the same storage logical page may be merged and processed in the next write request. That is, if some of the write data associated with the write request and the remaining data associated with the previous write request stored in the data I / O queue 126 are for the same storage logical page, they are merged, and the complete storage logical page for the merged logical page is merged. Whether or not can be distinguished.

In addition, if the host controller 124 is able to transfer a complete storage logical page immediately or if it is difficult to expect data that is the same storage logical page as the residual data in the next write request, then the write data is sent to the storage device 140. Can be. In one embodiment, host controller 124 corresponds to a complete storage logical page when the write data associated with the write request includes a complete storage logical page, so that the data corresponding to the complete storage logical page does not have to wait for the next write request. The data may be immediately transmitted to the storage device 140 without storing the data in the data input / output queue 126. In one embodiment, the host controller 124 may, in the next write request, if the write data does not contain data from the storage logical page corresponding to the residual data associated with the previous write request stored in the data input / output queue 126. Since it is difficult to expect data that is the same storage logical page as the residual data, the residual data may be transmitted to the storage device 140 as an incomplete storage logical page. In one embodiment, the host controller 124 may re-merge the data corresponding to the complete storage logical page into one I / O request and send it to the storage device 140.

In addition, when there is a flush request (hereinafter, referred to as a flush request) of the file system 122 to complete the processing of the requested write data, the host controller 124 may enter the current data input / output queue 126. All stored data may be transmitted to the storage device 140.

The host controller 124 should be interpreted to include all programs that process the transfer request in units of storage logical pages using the data input / output queue 126 between the file system 122 and the storage device 140. For example, the host controller 124 may be implemented in a device driver, a flash translation layer included in the device driver, or a program having a function similar to the flash translation layer. In an embodiment, the host controller 124 including the data input / output queue 126 may be implemented in a device driver that does not support the host flash translation layer function so that the transfer request may be performed in units of storage logical pages. According to an embodiment, the host controller 124 including the data input / output queue 126 may be implemented in the host flash translation layer included in the device driver so that the transfer request may be performed in units of storage logical pages.

Although not shown in FIG. 1, the host device 120 may further include other hardware or software components in addition to the file system 122 and the host controller 124, and the host controller 124 may also include data. In addition to the input / output queue 126 may further include other hardware or software components.

2A is a diagram illustrating an example in which a host controller operates in response to a write request of a file system in a conventional storage system, and FIG. 2B is a diagram illustrating an operation of the host controller in response to a write request of a file system in the storage system of FIG. 1. It is a figure which shows an example.

2A and 2B, effects of the write data transfer method of the host controller 124 are illustrated. Specifically, FIG. 2A illustrates a method in which a host controller 24 that does not have a data input / output queue processes a write request of the file system 22, and FIG. 2B illustrates a host controller (using the data input / output queue 126). 124 illustrates the manner in which the file system 122 handles the write request. 2A and 2B illustrate the case where the size of the storage logical page is four times the size of the host logical page. For example, the size of the host logical page may be 4 KB, and the size of the storage logical page may be 16 KB. In FIGS. 2A and 2B, the file system writes consecutive write requests for the storage logical pages L0, L1, and L2 to the first write request (L0-0, L0-1) and the second write request (L0-2, L0-3). , L1-0, L1-1, L1-2, L1-3, L2-0, L2-1), and the third write request (L2-2, L2-3) are divided into three requests . (Where Lx represents the storage logical page number and Lx-y represents the yth host logical page of the storage logical page Lx).

As shown in FIG. 2A, the host controller 24 having no conventional data input / output queue receives a write request from the file system 22 and immediately transfers write data to the storage device 40. (In FIG. 2A, WR1 represents write data received from the file system 22 and DT1 represents write data transferred to the storage device 40.) For example, the host logical page from the file system 22 is represented. When the first write request (L0-0, L0-1), which is twice the size, occurs, the write data (L0) is transmitted to the storage device 40, even though it is an incomplete storage page, and then the host logic If a second write request (L0-2, L0-3, L1-0, L1-1, L1-2, L1-3, L2-0, L2-1) that is eight times the page size occurs, the host The controller 24 transmits write data L0, L1, and L2 for each storage logical page to the storage device 40, respectively. Subsequently, when a third write request (L2-2, L2-3), which is twice the size of the host logical page, occurs, the write data (L2) corresponding to the same storage logical page as the previously transmitted write data is transmitted. do. In the above example, the host controller 24 sends two write requests for the storage logical page L0 and two write requests for L2. As described above, in the case of using the host controller 24 having no conventional data input / output queue, a write to the same storage logical page is transmitted in a plurality of write requests for a write request received from the file system 22. Frequently, this degrades the overall performance of the storage system and shortens its lifespan.

On the other hand, as shown in FIG. 2B, the host controller 124 including the data input / output queue 126 writes a plurality of write requests to the same storage logical page for a write request received from the file system 122. Can be prevented. In detail, the host controller 124 receives a write request from the file system 122 and stores write data associated with the write request in the data input / output queue 126. The controller 124 divides all data stored in the data input / output queue 126 into complete storage logical pages and incomplete storage logical pages so that data corresponding to the complete storage logical pages is transferred to the storage device 140 and is incomplete. Data corresponding to the storage logical page remains in the data input / output queue 126. 2B, WR2 represents write data received from the file system 122, DT2 represents write data transmitted to the storage device 140, and RD represents data corresponding to a complete storage logical page. The data stored in the data input / output queue 126 after being transmitted to the data input / output queue 126 is shown.) As described above, a case in which the host controller 124 uses the data input / output queue 126 is illustrated in FIGS. 2A and 2B. In a situation (that is, when the size of the storage logical page is four times the size of the host logical page), the first write request (L0-0, L0-1), which is twice the size of the host logical page, occurs from the file system 122. Host controller 124 stores data associated with the first write request in data input / output queue 126, and there is no complete storage logical page in data input / output queue 126. Therefore it does not transmit the data to the storage device 140. This is because data corresponding to an incomplete storage logical page is not transmitted to the storage device 140 until a write request of related data is completed to process data corresponding to the same storage logical page at a time. Subsequently, second write requests L0-2, L0-3, L1-0, L1-1, L1-2, L1-3, L2-0, and L2-1, which are eight times the size of the host logical page size, occur. In this case, the host controller 124 stores the write data associated with the second write request in the data input / output queue 126, and since the complete storage logical pages L0 and L1 exist in the data input / output queue 126, the storage device ( At 140, data L0 and L1 corresponding to the complete storage logical page are dequeueed from the data input / output queue 126 and transmitted. At this time, the host controller 124 may re-merge the data (L0, L1) corresponding to the complete storage logical page and transmit it as one I / O request. Subsequently, when the third write request (L2-2, L2-3), which is twice the size of the host logical page size, occurs, the host controller 124 writes data associated with the third write request to the data input / output queue 126. And a complete storage logical page (L2) exists in the data input / output queue (126), so that the data (L2) corresponding to the complete storage logical page in the storage device 140 is taken from the data input / output queue (126) and transmitted. . As described above, when the host controller 124 uses the data input / output queue 126, the number of write data transfers is reduced (in the above example, the number of write data transfers in FIG. 2A is five times, but is reduced three or two times in FIG. 2B). By avoiding duplicate write requests for the same logical page in successive writes of the file system, the overall performance of the storage system 100 can be improved and its life can be extended. In addition, there is an advantage of reducing the number of full merges during garbage collection processing in the flash conversion layer.

Referring to FIG. 3, when a flush request is received from the file system 122, the host controller 124 processes the process.

Specifically, in FIG. 3, the host controller 124 receives a write request from the file system 122, stores write data related to the write request in the data input / output queue 126, and stores data stored in the data input / output queue 126. In the case of including a complete storage logical page or a flush request from the file system, data is transmitted to the storage device 140. In FIG. 3, WR2 represents write data received from the file system 122, DT2 represents write data transmitted to the storage device 140, and RD transfers write data to the storage device 140. Data stored in the data I / O queue 126. For example, assuming that the size of the storage logical page is four times the size of the host logical page, the file system 122 has twice the size of the host logical page. When the first write requests L0-0 and L0-1 having sizes are generated, the host controller 124 stores the write data associated with the first write request in the data input / output queue 126 and the data input / output queue 126. ) Does not transmit data to the storage device 140 because no complete storage logical page exists. Subsequently, second write requests L0-2, L0-3, L1-0, L1-1, L1-2, L1-3, L2-0, and L2-1, which are eight times the size of the host logical page size, occur. In this case, the host controller 124 stores the write data associated with the second write request in the data input / output queue 126, and since the complete storage logical pages L0 and L1 exist in the data input / output queue 126, the storage device ( Data L0 and L1 corresponding to the complete storage logical page at 140 are taken from the data input / output queue 126 and transmitted. At this time, the host controller 124 may re-merge the data (L0, L1) corresponding to the complete storage logical page and transmit it as one I / O request. Subsequently, when the host controller 124 receives a flush request from the file system 122, the host controller 124 transfers all data L2-0 and L2-1 stored in the data input / output queue 126 to the data input / output queue 126. ) Is transferred to the storage device 140. Subsequently, when a third write request (L3-0, L3-1), which is twice the size of the host logical page size, occurs, the host controller 124 writes data associated with the third write request to the data input / output queue 126. And write data is not transmitted to the storage device 140 because a complete storage logical page does not exist in the data input / output queue 126. Here, when the flush request occurs, the file system 122 makes a third write request, even though there is space available for the storage logical page requested in the second write request (L2-2, L2-3). An example of skipping and using storage logical pages is shown. This is to avoid the write request for the remaining space because a plurality of write requests are sent to the corresponding storage logical page, which may cause a full merge in the storage controller.

Referring to FIG. 4, when a new write request is received from the file system (Step S110), the host controller may store write data associated with the new write request in the data input / output queue (Step S130). All data stored in the data I / O queue (including residual data related to previous write requests stored in the data I / O queue) may be divided into complete storage logical pages and incomplete storage logical pages (Step S150). In this case, the host controller transmits data corresponding to a complete storage logical page to the storage device in response to the write request (Step S170), while inputting / outputting data corresponding to an incomplete storage logical page in response to the write request. Can remain in the queue (Step S190).

Specifically, the file system reads data in block units and makes a new write request (Step S110), and the host controller processes the write data related to the write request of the file system as the host logical page size and stores it in the data input / output queue (Step S130). )can do. According to an embodiment, when the write data includes a complete storage logical page, the host controller may transmit the data corresponding to the complete storage logical page to the storage device without storing the data in the data input / output queue (Step S112). The data corresponding to the complete storage logical page does not need to wait for the next write request, so performance can be improved by transferring it to the storage device rather than storing it in a data I / O queue. In one embodiment, it is determined whether the write data associated with the new write request includes data of the storage logical page corresponding to the residual data associated with the previous write request (Step S114), and if not, the same as the residual data in the next write request. Since data that is a storage logical page is difficult to expect, residual data may be transferred to the storage device (Step S116). Here, determining whether the write data associated with the new write request includes data of the storage logical page corresponding to the residual data associated with the previous write request (Step S114) and transmitting the residual data to the storage device (Step S116). Regardless of the step and order of storing the data in the data input / output queue (Step S130), it may be determined and transmitted. The host controller may divide all data stored in the data I / O queue into a complete storage logical page and an incomplete storage logical page (Step S150). In one embodiment, whether or not the complete storage logical page is a complete storage logical page if it contains data corresponding to the last portion of the storage logical page, otherwise it may be determined as an incomplete storage logical page. In one embodiment, whether the data stored in the data I / O queue is a complete storage logical page compares the storage logical page size with the size of the data that is determined to correspond to the same storage logical page within the data I / O queue. In the same case, it may be determined as a complete storage logical page, and when the size of data determined to correspond to the same storage logical page in the data I / O queue is smaller than the storage logical page size, it may be determined as an incomplete storage logical page. Thereafter, the host controller may take data corresponding to a complete storage logical page from the data input / output queue and transmit the data to the storage device (Step S170). In one embodiment, the host controller may re-merge the data corresponding to the complete storage logical page into one I / O request and then send it to the storage device. On the other hand, the host controller may leave data corresponding to an incomplete storage logical page in a data input / output queue (Step S190). Meanwhile, according to an embodiment, when a flush request from a file system comes into the host controller, the host controller may transmit all data stored in the data input / output queue to the storage device to process data related to all write requests.

As mentioned above, the storage system and the write data processing method thereof according to the embodiments of the present invention have been described with reference to the drawings, but the above description is exemplary, and the general knowledge in the technical field of the present invention is not limited to the technical spirit of the present invention. It may be modified and changed by those who have.

The present invention can be applied to a storage system having a storage device (ie, a flash memory device). For example, the present invention can be applied to a storage system having a solid state drive (SSD), an embedded multimedia card (EMMC), a secure digital card (SD card), and the like.

Although the above has been described with reference to exemplary embodiments of the present invention, those skilled in the art may vary the present invention without departing from the spirit and scope of the present invention as set forth in the claims below. It will be understood that modifications and changes can be made.

* Explanation of the sign

100: storage system 120: host device

122: file system 124: host controller

126: data input / output queue 140: storage device

142: storage controller 146: flash memory

Claims

When a file system makes a write request, storing write data associated with the write request in a data input / output queue;

Dividing all data stored in the data input / output queue into a full storage logical page and an incomplete storage logical page;

Transmitting data corresponding to the complete storage logical page to a storage device in response to the write request; And

And remaining data corresponding to the incomplete storage logical page in the data input / output queue in response to the write request.
The storage system of claim 1, wherein when the write data includes the complete storage logical page, data corresponding to the complete storage logical page is transmitted to the storage device without being stored in the data input / output queue. How to handle write data.
The residual data of claim 1, wherein the remaining data is transmitted to the storage device when the write data does not include data of a storage logical page corresponding to residual data associated with a previous write request stored in the data input / output queue. Write data processing method of a storage system, characterized in that.
The method of claim 1, wherein the complete storage logical page and the incomplete storage logical page are classified based on a storage logical page size set in the storage device.
The method of claim 1,

And transmitting all data stored in the data input / output queue to the storage device when the file system requests a flush.
A storage device having a storage controller to control the flash memory using at least one flash memory and a storage flash translation layer; And

A host device having a file system and a host controller for interacting with the storage controller,

When the file system makes a write request, the host controller stores the write data associated with the write request in a data input / output queue, and converts all data stored in the data input / output queue into a complete storage logical page and an incomplete storage logical page. And classifying and transmitting data corresponding to the complete storage logical page to the storage device.
The data storage device of claim 6, wherein when the write data includes the complete storage logical page, the host controller transmits data corresponding to the complete storage logical page to the storage device without storing the data in the data input / output queue. Storage system.
The storage device of claim 6, wherein the host controller is configured to store the residual data when the write data does not include data of a storage logical page corresponding to residual data associated with a previous write request stored in the data input / output queue. Storage system, characterized in that the transmission to the device.
7. The storage system of claim 6, wherein the host controller classifies the complete storage logical page and the incomplete storage logical page based on a storage logical page size set in the storage device.
The storage system of claim 6, wherein the host controller transmits all data stored in the data input / output queue to the storage device when the file system requests a flush.